In a study published in 1992, the communication researcher Donald Rubin played American undergraduates a short recorded lecture. The voice belonged to a native speaker of English raised in central Ohio. While they listened, some students were shown a photograph of a white woman, presented as the lecturer, and others a photograph of a Chinese woman. Students in the second group were more likely to report that the lecturer spoke with a foreign accent, and they scored lower on a test of how much of the lecture they had understood. Nothing about the sound had changed. The accent they heard had been supplied by the photograph.
The study belonged to a particular moment. American universities had been hearing complaints from undergraduates about international teaching assistants, and many institutions responded, reasonably enough, by testing the spoken English of the assistants. Rubin’s experiment suggested, with a straight face, that some of the testing might profitably have been aimed at the other half of the room. I find that funny in the way good experiments about ourselves often are, and sobering in the same breath.
It is also easy to make the study say more than it does. It was one experiment, with one recorded voice, one kind of audience and one manipulation, carried out with students who had their own reasons to feel strongly about who taught them. It showed that a listener’s expectation can alter both the accent heard and the comprehension measured. It did not show how large that effect is among listeners in general, or that every listener who struggles with an unfamiliar voice is inventing the difficulty. Rubin and later collaborators gave the phenomenon a name, reverse linguistic stereotyping, in which a stereotype about the speaker shapes what the listener believes he hears. The name is useful precisely because it marks one mechanism among several. Some difficulty is real, some is supplied, and in an ordinary conversation nobody can see the photograph.
An accent arrives before the story that explains it. Mine belongs to an upbringing in Amritsar, with Punjabi as a first language, and to a medical education conducted in English, training in Arkansas and California, and practice in California since 2010. A voice keeps some record of that movement, even when it is saying something as ordinary as your blood pressure looks better today. Punjabi has a word for this that is wider than the English one: ਲਹਿਜਾ, lahijā, which covers accent but also tone and manner, the way a sentence is carried as well as the way its vowels are shaped. An accent in that sense is not only a set of sounds. It is a manner of arriving in a room.
Everyone has one. The phrase “no accent” usually means “an accent like mine,” spoken by whoever happens to be listening, and I say this with sympathy for the people who use it. It is very comfortable to assume that accents are something other people have. I suspect most of us, myself included, slip into that assumption whenever we are the ones who belong in the room, and climb out of it only when we are not.
Three Measures #
The most useful thing research has done for this subject is to separate words that ordinary speech runs together. In 1995 Murray Munro and Tracey Derwing reported, in Language Learning, a study of speech by adult learners of English in which listeners judged three things separately. The first was accentedness: how different the speech sounded from the listener’s own norm. The second was comprehensibility: how easy or difficult the listener felt the speech was to understand. The third was intelligibility: how much of it the listener actually understood, measured by asking listeners to write down what they had heard rather than to report an impression. The three measures were related, but they came apart in ways that matter. Speech could be judged heavily accented and still be transcribed almost perfectly. How foreign a voice sounded turned out to be an unreliable guide to how much of it had been understood. Munro and Derwing restated the point in a retrospective in 2020, a quarter of a century on, and it has held up well enough to become a foundation of the field.
The distinction between the second and third measures is the one a clinic most needs. Comprehensibility is a feeling, the sense of effort. Intelligibility is an outcome, the words correctly received. A listener can work hard and understand everything; a listener can feel entirely at ease and understand less than she thinks. The feeling is genuine information about the encounter, since effortful listening is tiring and a tired listener eventually misses things. But it is information about the listener’s experience, not a certificate of what was received.
None of the three asks whether the message was any good. Consider two hypothetical instructions about a new medicine. One is spoken in an accent the listener finds unfamiliar: Take one tablet in the morning with breakfast, never at night. The listener has to work a little, and transcribes every word. The other is spoken in the listener’s own accent, with perfect ease, and consists of three words familiar from a thousand pharmacy labels: Take as directed. The second scores beautifully on accentedness and comprehensibility, and conveys nothing at all. It is native speech in its purest administrative form, and it has sent more than one person home to guess. A voice can sound unfamiliar while conveying its meaning accurately. It can sound entirely familiar while conveying a poor explanation with great confidence.
Kazuo Ishiguro built a whole novel on the second possibility. Stevens, the butler who narrates The Remains of the Day (1989), speaks an English so polished that it has become a means of not saying things. His diction is impeccable and his account is evasive, and the reader learns to listen beneath the polish for what the voice has been trained to leave out. Restraint in Ishiguro is a form of meaning. In a clinic, polish without content is just a pleasant noise, and the ease it produces in the listener can be the most misleading thing in the room.
What Clarity Owes #
None of this makes clarity optional. In medicine there is no useful romance in being misunderstood. A patient who leaves the room with the wrong idea about a medicine or a warning sign has been let down, whatever the reason, and the clinician’s duty to prevent that does not depend on whose accent is involved. Repeating a sentence, choosing plainer words, slowing down, putting the one essential point on paper, stopping after each piece of a plan instead of delivering it in a single breath: these are acts of care, and they belong to every clinician, including the ones who sound exactly like their patients.
They also have to be checked rather than assumed. Plain language is a practice, not a property one possesses. A physician who believes he speaks plainly has a hypothesis about himself, and it can be tested only by what the patient is able to do with the words afterwards.
Those obligations do not require believing that only one kind of voice sounds competent, and the distinction matters to me. Adjusting how I speak so that a particular person can follow me is respect for that person. It may mean slowing down, choosing a different word, pronouncing a drug name the way the pharmacy label will spell it, or saying the number of tablets twice in different words. Trying to erase every audible sign of where I come from is a different undertaking. The first improves a conversation. The second turns every conversation into a continuous audition, and it is tiring to imagine that every vowel must submit supporting documents.
The line between them is not always sharp, and it should not be drawn self-servingly. If a particular sound reliably causes a particular word to be misheard, and the word matters, changing the sound is clarity, not surrender. People who work in more than one language make such changes all the time, often without noticing. What I resist is the idea that the measure of a voice is its distance from some other voice, rather than whether the person listening has understood.
Mrs Higgins’s At-Home #
English literature’s great comedy of accent is Bernard Shaw’s Pygmalion, first performed in 1913 and in London in 1914, and its preface opens with the sentence that has made it permanently quotable: “It is impossible for an Englishman to open his mouth without making some other Englishman hate or despise him.” Shaw’s target is not the speaker. It is a society that has turned vowels into a sorting machine for class, and a professor of phonetics, Henry Higgins, who can place a stranger within a few streets by her speech and who treats this skill as a licence for contempt.
The play’s most instructive scene comes before the transformation is complete. It is the first public test, in the third act, at Mrs Higgins’s at-home day. Eliza Doolittle, the flower girl whose vowels Higgins has been rebuilding, arrives speaking with flawless precision. Then, invited to make conversation, she delivers in those perfect vowels the history of her aunt’s death. “My aunt died of influenza: so they said,” she announces, and adds, in the same tragic tone and the most refined accent in the room, “it’s my belief they done the old woman in.” She explains her reasons. The aunt had survived diphtheria the year before, turned blue with it, and been revived by Eliza’s father ladling gin down her throat. Why should a woman of such strength die of influenza? And what had become of the new straw hat that was due to come to Eliza?
The guests do not hear a flower girl. They hear the right vowels, and so they classify everything else accordingly. Higgins smooths things over by explaining that this is the new small talk, and the younger guests are delighted: a young man among them is charmed, and his sister leaves practising the same daring vocabulary. The sound has decided the meaning. It is Rubin’s experiment run in reverse, nearly eighty years before Rubin: there the listeners’ expectation supplied an accent that was not present; here it removes a social origin that is entirely present, because the vowels say it cannot be.
A physician reads the scene with an additional pleasure. Eliza’s speech is, among other things, a clinical history, with an acute respiratory illness, a previous episode of diphtheria with cyanosis, a dubious resuscitation by spirits, and a family member’s suspicion of foul play. It is perfectly intelligible. Nobody in the room hears it as information, because the accent has filed it under manners. That is the subtler danger of accent judgments: not only that unfamiliar voices are heard as less competent, but that familiar ones are heard as saying something other than what they say.
Shaw also counts the cost of the transformation, which the musical version tends to sweeten. In the fourth act, the experiment won, Eliza asks the question no one has thought to answer: “What’s to become of me?” Her new voice has made her unfit for the flower market and has not provided her with a place anywhere else. An accent can be changed. A life is not thereby arranged. The comedy is aimed, throughout, at the men who thought the vowels were the whole of the problem.
The Language Becomes Mine #
The writers who did most to change how English sounds on the page were those who came to it from elsewhere and refused to apologise for what they brought. Raja Rao, in the foreword to his novel Kanthapura in 1938, wrote of the difficulty of conveying a spirit that is one’s own in a language that is not, and said that the method of expression would have to be a dialect of English that might one day prove “as distinctive and colourful as the Irish or the American.” The comparison is shrewd. Nobody now thinks that Irish or American English is a failed attempt at the English of London. They are Englishes, with their own histories and authority, and Rao was claiming the same standing, in advance, for an English shaped by Indian rhythms.
Chinua Achebe made a related argument in the 1960s, in his essay “The African Writer and the English Language.” He thought English could carry the weight of his African experience, but only as a new English, “still in full communion with its ancestral home,” altered to suit its new surroundings. The condition he attached is easily overlooked and seems to me the heart of the matter. Full communion means the new English remains intelligible to the old. Achebe was not claiming a right to be obscure. He was claiming the right to be himself and understood at once.
Ngũgĩ wa Thiong’o drew the opposite conclusion and left English for Gikuyu as the language of his fiction; his argument, and its bearing on the language of medicine, is the subject of another essay on this site. For the question of accent, the more useful voice is Kamala Das. In “An Introduction,” from her first collection, Summer in Calcutta (1965), the speaker answers everyone who has told her not to write in English because it is not her mother tongue. The language she speaks, she says, becomes hers, with its distortions and its oddities, and she describes it as “half English, half Indian, funny perhaps, but it is honest.” Every word of that line does work. Half English, half Indian declines the choice between imitation and rejection. Funny perhaps concedes, with a shrug, that the result may amuse the people who set the standard. Honest is the claim that matters: the language is true to the person speaking it.
These were arguments made on the page, where a writer has time, an editor and the chance to revise. A spoken accent makes the same claim without any of those advantages, in real time, in a room where someone has to understand an instruction correctly the first time. So the writers’ arguments transfer to the clinic only with Achebe’s condition attached. An accent can be honest in Das’s sense and still owe the listener clarity. The two obligations are not in tension. Honesty is the right to remain oneself; communion is the duty to remain intelligible; a good clinical voice keeps both.
Departing for Foreign #
Indian English has its own comic literature, and its greatest practitioner was Nissim Ezekiel, whose sequence of “Very Indian Poems in Indian English,” from the 1970s, is affectionate and merciless in about equal measure. In “Goodbye Party for Miss Pushpa T.S.” a speaker at a farewell explains that the honoured colleague “is departing for foreign / in two three days,” and continues in the same idiom, with present continuous verbs for everything, through a speech of increasing warmth and decreasing coherence. Readers laugh, and have argued for decades about what they are laughing at.
The argument is worth having because the poem can be read two ways. Read carelessly, it mocks a way of speaking, and a reader who already looks down on Indian English will find his view confirmed. Read with attention, the joke lands elsewhere: on the inflation of the occasion, on the speaker’s need to say something weighty about a colleague he barely knows, on the universal comedy of the farewell speech, which is no less absurd in any other English. Ezekiel, who wrote most of his poetry in an English as standard as any, knew exactly what he was risking. The poems are ventriloquism done from inside, and their final effect is closer to love than to contempt. Whether a joke about speech is cruel depends on where the laugh is aimed.
That seems to me the right rule for humour about accent in a clinic, where humour is often necessary and occasionally dangerous. It is safe when it is aimed at systems, at the photograph and the tests aimed at only one half of the room, at the pharmacy label and its three famous words. It is safe when it is aimed at oneself. It is not safe when it is aimed at the way a patient or a colleague speaks, however affectionate the intent, because the person on the receiving end cannot tell, any more than Ezekiel’s careless readers could, which way the laugh is facing.
On that rule, I am entitled to one joke at my own expense. I, too, departed for foreign. I have tried not to make a speech about it, though this essay may count against me. And I am as confident as anyone that my own vowels are neutral, which is to say, as confident as everyone is about their own vowels and with exactly as little evidence.
The Listener’s Minute #
The listener’s work is real, and there is evidence that it repays the effort quickly. In 2004 Constance Clarke and Merrill Garrett reported in the Journal of the Acoustical Society of America that listeners who were at first slower to process English spoken with an unfamiliar accent caught up within about a minute of exposure, until the gap between the accented speaker and a speaker of their own variety had largely closed. In 2018 a team led by Xin Xie revisited the finding and reported both that the rapid adaptation could be reproduced and that some of it carried over to a new speaker with the same accent. Ears adapt, and they adapt fast.
The finding is encouraging, and it is also easy to inflate into a moral rule: that any listener who has not adjusted within a minute has failed in fairness. The experiments do not support that. They were laboratory studies, in which listeners heard recorded sentences in quiet conditions and performed speeded tasks designed to measure how quickly speech was processed. The minute is a measure of processing speed for clear recorded speech under good conditions. It is not a promise about a crowded emergency department, a telephone line that compresses every consonant, a masked face, a listener whose hearing has been declining for a decade, or a frightened patient who has just heard a word she was dreading and stopped listening to whatever came after it.
A clinic is very often a poor laboratory. Hearing loss is common in older adults, and many people who have it do not know or do not say. Noise is ordinary. Fatigue and fear are ordinary. Each of these makes any speech harder to follow, accented or not, and together they can make an unfamiliar voice genuinely difficult without any prejudice being involved. Rubin’s photograph shows that expectation can do part of the hearing. It does not show that expectation does all of it.
What the research supports is more modest and more useful. Listeners can adapt, often faster than they expect to, and the adaptation is helped by attention and by the willingness not to decide in advance that the effort will be wasted. Speakers can make adaptation easier by pace, by pausing between ideas and by saying important things in more than one way. Neither party needs to settle in advance who is to blame for a difficulty before both begin to reduce it. Responsibility does not have to be allocated completely before communication improves.
An Unequal Share #
From the patient’s side of the desk, the argument so far can look like an unfair division of labour. A patient is not a participant in a listening experiment with a minute to spare. She has come because something is wrong; she may be in pain or afraid, and her attention is already spent on the thing she came about. She did not choose her physician’s voice, and in many places she has little choice of physician at all. To tell her that understanding is shared labour, and that she owes a fair hearing, can sound like asking the sick to do part of the doctor’s job. A patient who says she cannot understand her doctor may simply be right, and the right response is to make sure she can, not to speculate about her expectations.
I accept most of this. The shares are not equal, and they were never meant to be. The clinician is the professional, trained and paid to communicate, and holds most of the power in the room; the patient is unwell, and illness consumes the very attention that adaptation requires. So the burden lies more heavily on the person in the white coat than on the person in the gown, whatever their accents. If a patient cannot follow me, the first question is what I can change, and the question of what she might change comes much later, if it comes at all.
Where the listener’s share applies with full force is elsewhere. It applies to colleagues: the nurse taking a telephone call from a voice she has not heard before, the consultant listening to a handover, the committee interviewing a candidate. It applies to institutions, which collect judgments about voices in evaluations and complaints and let those judgments follow a person through a career. Rubin’s students were not patients. They were the kind of listener whose impressions become ratings, and ratings become decisions. And it applies to the physician himself, who listens to patients all day and whose expectations about their voices shape what he hears.
The asymmetry has one further consequence for institutions. When a complaint about a clinician’s accent arrives, it should be treated the way a good clinician treats a symptom: as real, and as not yet diagnosed. Sometimes the clinician needs help with pace or with a particular sound that keeps causing a particular word to be misheard. Sometimes the room is too loud, the telephone line too poor, the visit too short for anyone to be understood. Sometimes the complaint is the photograph. The complaint is data. Accepting its first explanation without examination, in either direction, is the institutional equivalent of a missed diagnosis.
Words Built to Be Heard #
Some of the most effective answers to mishearing have come from systems that declined to ask whose fault it was. Among the patient-safety goals it began setting for hospitals in 2003, the Joint Commission asked that a verbal or telephone order, and later a critical test result reported by telephone, be read back in full by the person who receives it. The rule applies to everyone, including two people with identical accents on a perfect line. It rests on the sound assumption that hearing fails routinely, and it builds the check into the exchange instead of waiting to discover the failure.
Aviation went further, and changed the words instead of the speakers. The international spelling alphabet adopted in the 1950s, in which letters are spoken as Alfa, Bravo, Charlie and so on, was chosen after testing with speakers of several languages, and some of its spellings were adjusted so that they would be pronounced as intended by people whose first language was not English: Alfa with an f, Juliett with two t‘s. Nobody in that history seems to have proposed that the world’s pilots adopt the vowels of a single English county. The solution was to find words that survive accents. Medicine has its own modest versions, in lists of drug names that sound alike and are known to be confused, and in the habit of spelling a name aloud or stating a number twice in different forms.
What these systems share is a refusal to treat understanding as a property of a voice. They treat it as an outcome of an exchange, to be verified, and they design the exchange so that verification is ordinary and blameless. That is the institutional form of the courtesy this essay keeps returning to: assume that mishearing will happen, make checking routine, and leave the question of fault aside.
Checking the Explanation #
The health-literacy literature has a name for the practice that turns all this into a habit: teach-back. The Agency for Healthcare Research and Quality includes it in its Health Literacy Universal Precautions Toolkit, and the idea is simple. After explaining something that matters, the clinician asks the patient to say in her own words what she will do. If the account is right, the explanation worked. If it is wrong or incomplete, the clinician explains again, differently, and checks again.
What I like about teach-back is where it places the burden. Done well, it avoids Do you understand?, which mostly invites a polite yes, and says something closer to I want to be sure I explained that clearly; can you tell me how you’ll take it? The test is of the explanation, not of the patient. That small grammatical courtesy has large consequences. It allows a patient who did not follow to say so without confessing to anything, and it allows a physician whose accent was a barrier to find out without anyone having to name it.
Teach-back also locates the problem, which repetition alone cannot. A request to repeat a sentence does not, by itself, identify the source of the difficulty. The wording may be unfamiliar. The room may be noisy. The listener may have missed the beginning, may have difficulty hearing, or may be interpreting the voice through an expectation formed before the sentence began. The speaker may need to change pace or pronunciation. Several of these conditions can coexist. Repeating exactly the same words more loudly may help in one situation and leave another untouched. Asking the listener to describe the plan reveals where the gap actually lies: whether a word was misheard, a step was missed, or the whole idea arrived intact except for the one detail that mattered.
The method has limits worth stating. It takes time, and time is the scarcest thing in many clinics. It can become a ritual, recited at the end of a visit to satisfy a checklist, in which case it checks nothing. It can feel like an examination to a patient if it is introduced carelessly. None of these is an argument against it. They are arguments for doing it as what it is meant to be, a check on the speaker’s work.
The Physician as Listener #
A physician in California spends at least as much of the working day listening to accents as producing one. English shaped by many first languages, English learned last year and English spoken for eighty years: an examination room is a meeting of speech histories, more than two if an interpreter or a relative is present, and the obligation runs in every direction. Whatever I ask of my listeners, I owe to my speakers.
That includes the honesty to admit when I have not understood. A physician who nods at a sentence he did not catch, because asking again feels awkward, has done exactly what a patient does when she says yes to Do you understand? The cure is the same. Let me make sure I have this right followed by a summary of what the patient said is teach-back turned around, and it does for the listener what teach-back does for the speaker. It checks the hearing instead of assuming it.
It also includes resisting two opposite errors of expectation. One is to assume that a patient who speaks with an unfamiliar accent understands less English than she does, and to simplify until the explanation becomes condescending. The other is to assume that a patient whose English sounds like one’s own has understood, because the ease of the exchange felt like understanding. Both errors confuse the sound of a voice with the state of a mind. The second is the harder to catch, since nothing in the conversation feels wrong.
Not every barrier is an accent, and a physician should not let the subject swallow the others. A patient who prefers another language is entitled to a qualified interpreter, a matter with its own law and its own difficulties. A patient with hearing loss needs a quieter room, a face that can be seen, a slower pace or a written summary, not a lecture on adaptation. A patient who is terrified needs the most important sentence again in ten minutes. When I ask a patient to repeat something, I want the request to sound like what it is, a request, and not a verdict on the way she speaks. I would like the same courtesy extended to every clinician, whatever the voice.
Machines That Listen #
Machines have a share in this too, and their history in it is mildly comic. Early dictation software asked each new user to read prepared passages aloud before it would take dictation, so that it could learn the speaker’s voice. It was the only listener I can think of that insisted on being read a story before it would pay attention. The systems now beginning to listen in examination rooms learned instead from enormous quantities of recorded speech, and whether they hear every voice equally well is a fair question to keep asking of them, as it is of people. A listener that has heard only one kind of English has not really learned English. It has learned an accent and mistaken it for the language.
The question deserves evidence rather than slogans. In 2020 a team at Stanford led by Allison Koenecke reported in the Proceedings of the National Academy of Sciences that five widely used commercial speech-recognition systems made nearly twice as many word errors, on average, when transcribing Black American speakers as white American speakers. That study concerned dialect, not foreign accent, and particular systems at a particular time, and it measured transcription, not consequences. Each of those qualifications matters. A disparity found in one set of systems does not prove the same disparity in another, and systems are retrained constantly. Accent strength, transcription error and clinical harm are three different measurements, and they are related only loosely: a system can garble a dozen filler words harmlessly and drop a single not with serious effect.
What follows is a demand for specific evidence, task by task. If a system is going to listen to clinical conversations and draft the record, the question that matters is how accurate it is for these speakers, in these rooms, on the words that change decisions, and whether anyone checks. Accuracy in general settles very little. The same Munro–Derwing distinction applies to machines as to people. A transcript that reads smoothly is the machine’s version of comprehensibility, the feeling of ease. Whether it has correctly captured what was said is intelligibility, and only checking can establish it.
A Voice of Its Own #
I want a voice that can be clear and still be mine. That seems a modest wish, yet it contains a negotiation between belonging and self-respect that I doubt is ever entirely settled. It also leaves room for change. A voice need not stay fixed to stay honest. People pick up words, rhythms and habits from the places and people around them, and a voice that has lived in more than one place is entitled to show it. Someone who chooses to soften an accent for reasons of his own has not betrayed anything; someone who keeps it has not failed at anything. Neither retention nor assimilation is a moral duty. Eliza’s question, what’s to become of me?, is a reminder that the person inside the voice has interests the vowels do not settle.
What does seem to me a duty is the thing both speaker and listener can actually do: make the meaning cross, and check that it has. The speaker owes clarity, pace and a willingness to say it differently. The listener owes attention, a fair hearing and the honesty to ask again. Both owe each other the assumption that a request to repeat is a request. The feeling of ease is useful information about an interaction and a poor certificate of the speaker’s knowledge. What matters is whether the account is accurate, understandable and responsive to the person who needs to use it.
Whose work is it, then, when a voice is hard to follow? The evidence supports a plain answer: the work is shared, unequally, and it can begin before anyone has settled the shares. If I listen only for familiarity, I may miss the meaning. If I listen with a little patience and then check what I heard, an unfamiliar voice becomes a particular person, often within the minute the laboratory measured and sometimes only after a good deal longer. Most conversations that matter last longer than a minute. It seems a small price to pay at the start of one, and a smaller one than the price of a plan misunderstood in a perfectly familiar accent.
Evidence and Editorial Note #
Documented findings. Public biography: Punjabi as first language; upbringing in Amritsar; medical education in English; training in Arkansas and California; practice in California since 2010. Rubin DL, “Nonlanguage factors affecting undergraduates’ judgments of nonnative English-speaking teaching assistants,” Research in Higher Education 33(4):511–531 (1992), as summarised in secondary accounts drawing on Rosina Lippi-Green’s English with an Accent; the term “reverse linguistic stereotyping” and the context of complaints about and testing of international teaching assistants are stated from standard sociolinguistic knowledge. Munro MJ, Derwing TM, “Foreign accent, comprehensibility, and intelligibility in the speech of second language learners,” Language Learning 45(1):73–97 (1995), and “Foreign accent, comprehensibility and intelligibility, redux,” Journal of Second Language Pronunciation 6(3) (2020). Clarke CM, Garrett MF, “Rapid adaptation to foreign-accented English,” Journal of the Acoustical Society of America 116(6):3647–3658 (2004); Xie X et al., “Rapid adaptation to foreign-accented speech and its transfer to an unfamiliar talker,” Journal of the Acoustical Society of America (2018); the description of their tasks as laboratory measures of processing speed with recorded speech is a general characterisation. The teach-back method as included in the Agency for Healthcare Research and Quality’s Health Literacy Universal Precautions Toolkit: retained in general terms; The Joint Commission’s National Patient Safety Goals from 2003 onward, requiring read-back of verbal and telephone orders and, later, of critical results reported by telephone, and the adoption in the 1950s of the international radiotelephony spelling alphabet with the spellings Alfa and Juliett: well-established institutional history, stated in general terms. Koenecke A et al., “Racial disparities in automated speech recognition,” Proceedings of the National Academy of Sciences (2020), described in general terms from well-established knowledge; no system is named. Literary sources: Bernard Shaw, Pygmalion (first performed 1913; London 1914), preface and Acts III–IV, quoted from the public-domain text; Raja Rao, foreword to Kanthapura (1938); Chinua Achebe, “The African Writer and the English Language” (1960s), paraphrased; Kamala Das, “An Introduction,” Summer in Calcutta (1965); Nissim Ezekiel, “Goodbye Party for Miss Pushpa T.S.,” from his “Very Indian Poems in Indian English”; Kazuo Ishiguro, The Remains of the Day (1989). Quotations from copyrighted works are single short lines; The Punjabi ਲਹਿਜਾ (lahijā), accent or manner of speaking: standard usage. Vendor representations. None. No speech-recognition or documentation product is named or assessed.
Recollection and tradition. The author’s languages and training are his published biography. No encounter, remark, misunderstanding or incident of bias involving the author’s speech is described or implied, and no claim is made about how the author pronounces any particular word. The two instructions compared under “Three Measures” are hypothetical. Remarks about listening in examination rooms are general descriptions of clinical practice, not accounts of particular patients or colleagues.
Analysis and forecast. The distinction between clarity as care and accent conformity as audition, the reading of Shaw’s at-home scene as Rubin’s experiment in reverse, the reading of Achebe’s “full communion” as an intelligibility condition, the rule about where humour is aimed, the reply to the objection that patients cannot share the work equally, the account of the listener’s share and its limits, and the remarks on listening machines are the author’s present-day interpretation as of November 6, 2026. Independence. This essay is not sponsored by any institution, company, health system or technology developer. It is personal reflection, not medical advice; readers seeking care should contact their own physician’s office.
About the Author #
Kanwar Partap Singh Gill, MD, traces his family roots to Khadur Sahib in Punjab’s Majha region and grew up in Amritsar. He attended Spring Dale School and completed pre-medical studies at Khalsa College in 1994–1996. He earned his MBBS from Sri Guru Ram Das Institute of Medical Sciences and Research in Amritsar in 2002.
He completed PGY-1 at the University of Arkansas for Medical Sciences in Little Rock in 2007–2008, and PGY-2 and PGY-3 with UCSF Fresno at Community Regional Medical Center from 2008 through 2010. He has practiced medicine in California since 2010.
His career has crossed four distinct information eras: predominantly paper-based medicine, early and fragmented electronic records, the integrated-EHR era, and the emerging age of generative and agentic artificial intelligence.